With the increasing demand for automation in logistics due to a declining labor force, there is a growing need for advanced object recognition techniques. While vision-based recognition remains prevalent, its reduces accuracy when objects are visually similar or partially occluded. To address these limitations, we propose a non-destructive and force-sensorless method for estimating mechanical properties, such as spring constant, damping, and mass, by modeling objects as mass-spring-damper systems. Our approach combines a Reaction Force Observer (RFOB), which estimates contact force solely from actuator signals, with Recursive Least Squares (RLS) for parameter estimation. Furthermore, two-degree-of-freedom vibrational model distinguishes between surface and internal mechanical characteristics through a combination of quasi-static pressing and micro-vibration excitation. This enables the evaluation of structural non-uniformity in soft materials such as fruits and processed food items, which are often difficult to assess using visual sensing alone. The proposed method minimizes the need for additional force sensors, offering a cost-effective and flexible solution for robotic systems. This technique contributes to enhanced grasping and object recognition strategies, allowing robots to infer the internal properties of soft and complex objects during manipulation.
Yokoyama et al. (Thu,) studied this question.